Hello Experts,
My outcome is a scale variable (3 separate domains-continuous)
Covariates are mixed (categorical and continuous)
I am developing a 3 level model (care aides, unit and facility), unfortunately i found a warning message (iteration was terminated but convergence has not been achieved. the mixed procedure continues despite this warning. subsequent results produced are based on the last iteration. validity of the model fit is uncertain). In this scenario, R2 and ICC was very poor.
Now, I moved to 2 level model (care aides, Unit). here the same problem, I can not put my important covariates in this model. If i exclude some of my important covariates, there are no warning message.
Same models works in STATA but SPSS showed warning message.
Q1> Is that the limitations for SPSS?
Q2> Is there any options that If I can change, the problem will be resolved?
If you are using the default settings in both packages, the SPSS MIXED command is using restricted maximum likelihood (REML), whereas Stata's -mixed- command is using maximum likelihood estimation (option ml). That might make a difference in what you are seeing. HTH.
Good afternoon Ashikur Rahman, Puzzling no doubt for you. Covariates are mixed (categorical and continuous) I am developing a 3 level model (care aides, unit & facility). Is it possible for the continuous covariate maybe to categorise it into 2 levels rather than have it as continuous & would that change problem you are having.? I don't know that this is the answer.
Deborah J Hilton, carving a good quantitative variable into categories is almost always a bad idea. Please see this list of problems associated with that practice:
- https://discourse.datamethods.org/t/categorizing-continuous-variables/3402
HTH.
Thankyou Mr/Dr Weaver, I mentioned it as in my dissertation that was one method recommended to me by a statistician for dealing with glucose values to use a cut point for categorisation. I will read your link now or later. TY